AI Power: Complete AI Pack <= 1.8.96 - Authenticated (Admin+) PHP Object Injection via wpaicg_export_prompts
Strategic Overview
- Status
- Patched in 1.8.97
- Affected Plugin
- AI Puffer – Chat. Create. Automate. (formerly AI Power)
- Affected Version
<= 1.8.96- CVSS
- 7.2High
- Weakness type
- CWE-502 · Deserialization of Untrusted Data
- CVE
CVE-2025-0428
At a glance
CVE-2025-0428 is a high-severity Deserialization of Untrusted Data vulnerability in the AI Puffer WordPress plugin, affecting versions <= 1.8.96. It carries a CVSS score of 7.2 (reachable over the network; low attack complexity; high confidentiality, integrity, availability impact). Exploitation requires an authenticated account at Admin level or above. The issue is fixed in version 1.8.97; sites on affected versions should update now. Disclosed January 2025, reported by Tran Anh Duc.
Vulnerability Overview
The "AI Power: Complete AI Pack" plugin for WordPress is vulnerable to PHP Object Injection in versions up to, and including, 1.8.96 via deserialization of untrusted input from the $form['post_content'] variable through the wpaicg_export_prompts function. This allows authenticated attackers, with administrative privileges, to inject a PHP Object. No POP chain is present in the vulnerable plugin. If a POP chain is present via an additional plugin or theme installed on the target system, it could allow the attacker to delete arbitrary files, retrieve sensitive data, or execute code.
Technical Analysis
The vector marks this flaw as remotely reachable over the network, with low attack complexity — no special timing or configuration is needed, and no interaction from a victim user. A successful exploit has high impact on confidentiality, integrity, availability — full site compromise territory.
CWE-502: Deserialization of Untrusted Data
AI Puffer <= 1.8.96 carries this weakness at post_content, and reaching it takes an account at Admin level or above. Deserialization of untrusted data means the application reconstructs objects from input a caller controls, letting the attacker choose which classes get instantiated and what their properties hold.
A workable chain turns deserialization into file writes, deletions or code execution, depending on what the loaded classes do in their magic methods. For AI Puffer the fix is 1.8.97: builds <= 1.8.96 are affected, anything from 1.8.97 onward is not.
Remediation
Update to version 1.8.97, or a newer patched version
How does WordSec protect against this?
An attacker needs Admin access first, so the firewall sees the attempt as traffic from a logged-in account: WordSec's web application firewall inspects request payloads before WordPress loads them. Classes like this one can leave something behind, so the staged malware scanner is the second half: it looks for dropped files and modified code rather than for the request that created them. None of that substitutes for the fix: AI Puffer 1.8.97 closes this, and updating the plugin is the step that ends it.
- Firewall
- Scanner
- Alerts
External References
Related records
Other vulnerabilities in AI Puffer – Chat. Create. Automate. (formerly AI Power)
- 9.8CVE-2024-10392: AI Power: Complete AI Pack Arbitrary File Upload
CVE-2024-10392 - 8.1CVE-2023-0405: GPT AI Power <= 1.4.37 Missing Authorization
CVE-2023-0405 - 6.4CVE-2024-37465: GPT3 AI Content Writer <= 1.8.66 Stored XSS
CVE-2024-37465 - 6.3CVE-2024-13361: AI Power Authenticated (Subscriber+) Arbitrary
CVE-2024-13361 - 6.1CVE-2024-13362: Freemius <= 2.10.1 Reflected DOM-Based XSS
CVE-2024-13362 - 6.1CVE-2023-33999: Freemius SDK <= 2.5.9 Reflected Cross-Site Scripting
CVE-2023-33999 - 5.4CVE-2024-13360: AI Power: Complete AI Pack <= 1.8.96 SSRF
CVE-2024-13360 - 5.3CVE-2023-51527: AI Power: Complete AI Pack Sensitive Data Exposure
CVE-2023-51527
Vulnerability data © Defiant, Inc., provided under the Wordfence Intelligence T&C